A pretest-posttest control group design was used with 20 second- and third-grade teachers to answer questions related to the effectiveness of microteaching as a teacher training procedure. The group was divided into two experimental groups by random assignment. One group of 10 received microteaching training in a specific teaching skill. The other group viewed perceptual (videotaped) models as training in the same teaching skill. The Verbal Behavior Q Sort (VBQS) was used to define and measure teachers' intentions and perceptions of verbal behavior. Flanders Interaction Analysis Instrument (FIAI) was used to measure teachers' actual classroom behavior. The FIAI data and the VBQS data were analyzed for significant changes within and also between experimental groups. The analyses allowed the following conclusions to be made: (1) elementary teachers trained by microteaching technique or by viewing perceptual models made significant changes in their intentions and perceptions of classroom verbal behavior; and (2) in their intention and self-perception of using criticism in classroom verbal behavior, teachers trained through microteaching differed significantly from those trained with perceptual models. (Author/EB)